1. Documentation

Domain knowledge (1.1)

Since launching in May 2013 with 6000 bikes and 5000 members, Citi Bike is now the United State’s largest bike share program.

By July 2020, their customers had taken over 100 million trips, and today they have 20,000 bikes across 1,300 stations in Manhattan, Brooklyn, Queens, the Bronx, and Jersey City.



They offer a range of passes to cyclists who wish to enjoy quick trips in a fun and affordable way, including the annual pass which allows you unlimited 45 minutes rentals, or day and single trip passes which each allow for 30 minutes of cycling.

Should you go over your 30 or 45 minutes, there is a small charge per minute until you return and dock your bike at one of their stations (you can return to any station that conveniences you).



This data covers the Jersey City area and contains information on:

  • A sample from NYC Citi Bike usage of 10 bikes throughout 2018
  • Event data on each trip, including the trip’s start and end times and locations
  • The customer’s gender and birth year
  • The type of trip:
    • “Customers” are those who have purchased a 1 or 3 day pass, and
    • “Subscribers” are those who have purchased an annual subscription


This dataset contains information from:

    930 women                3,069 men

this image is from: “https://www.flaticon.com/free-icons/woman” title=“woman icons” - Woman icons created by Freepik - Flaticon this image is from: “https://www.flaticon.com/free-icons/man” title=“man icons”>Man icons created by Freepik - Flaticon>


3,953 "subscribers"       315 "customers"

this image is from: “https://www.flaticon.com/free-icons/subscribe” title=“subscribe icons” - Subscribe icons created by Freepik - Flaticon this image is from: “https://www.flaticon.com/free-icons/day” title=“day icons” Day icons created by Freepik - Flaticon


These numbers don’t match the total rows (N = 4268) because there were a few odd entries that were removed for analysis purposes. This is discussed later in this project.


For customer protection when considering data ethics:

  • This dataset is provided according to the Citi Bike Data Use Policy.
  • Citi Bike provides monthly reports to the NYC Department of Transportation with the intention of being open and transparent with their customers and the residents of the City of New York.
  • The data is minimised and anonymised to protect customer identity and personal behaviours:
    • Though the customer’s birth year and gender is recorded in the data set, no personal identifiers such as name, address, or contact details are present.
    • Each journey is recorded under the bike ID rather than a customer ID.
    • Though it is noted whether a customer is a “subscriber” or “customer”, there are no records of payment or any bank details.


Jersey City



Jersey City is separated from New York by the Hudson River with ferries and trains connecting the two; especially handy for commuters who might work in the busy business and financial districts.

With a lively downtown area, visitors also come to Jersey City to explore Liberty State Park for panoramic views of the Manhattan Skyline, the Ellis Island Immigration Station, and the Statue of Liberty.

Citi Bike suggest a popular cycling route that takes in all of these views (across a few hires, but doable in a day).


Image from: https://www.triptipedia.com/tip/8qdQMl2/3-amazing-manhattan-views-from-jersey-city


Business requirements (1.2)

Having researched the company as well as Jersey City, I plan to decipher the most popular areas for bike hire and return, as well as the most popular days of the week, to see if there are possible commuter/visitor clues that can be further looked into. Additionally, gender and age would be of interest with thoughts of targeting marketing at the most popular demographics.

With a view to providing actionable insights with the overall objective of increasing bike hires, this report follows the below steps:

  • Preparing (wranggling) and investigating the data set for analysis and visualisation
  • Identifying and investigating variables of interest and their relationships with each other with the use of visualisations
    • Focusing on patterns of bike hire over time
    • Considering whether patterns differ between rider demographics
    • Viewing the geographical spread of starting stations for bike hires
  • Ensuring that insights convey a clear and accurate message for the business


Business processes and data flow (1.3)

This diagram was created using draw.io at https://app.diagrams.net


Data visualisation as a tool for decision-making (1.4)

With consideration of user demographics as well as stations that are used most often for both rental and return, Citi Bikes can tailor their marketing to the most appropriate potential customers with the intention of increasing their bike hires throughout the year.

Further, they could look into stations that may require more or fewer bikes due to user numbers. This would be especially useful at this time as it is reported that the system has become more popular throughout the COVID pandemic and more people are returning to their offices having been working from home.


Data types (1.5)

Across the 12 variables, the data types are as follows:

Factor

  • A factor is a categorical variable that stores both string and integer data values as levels.
    • bike_id
    • start_station and end_station
    • type
    • and gender are all factor variables.

Datetime

  • Datetime variables can store calendar dates and times to the microsecond.
    • start_time and stop_time are datetime variables.

Double

  • This is a numeric data type that holds real numbers with at least 2 decimal places, e.g. 2.00
    • start_lat and start_long
    • end_lat and end_long
    • and birth_year are all double variables.


Data quality and data bias (1.6)

There are far more men (n = 3069) than women (n = 930), and 269 customers gender is recorded as “unknown”. This is not a huge concern as it may give insight into future marketing. Having looked into those with gender as “unknown” however, 243 out of 267 customers were recorded as being 49 years old. Further investigation into those aged 49 years revealed that 180 were recorded as a “customer” (1 or 3 day pass users) out of a total of 315 “customer” records. Something may have happened in the retrieval of this data as these figures seem unlikely. When considering differences between gender usage, those with “unknown” were therefore removed from the data set before analysis.

A couple of users are recorded as being born in the 1880s. They are both subscribers but also have no gender attached to themselves and only have one record each. I therefore removed these from the data set so that they do not skew the data.

Following just 10 bikes may not present a big enough picture - especially considering there are 82 stations in Jersey City and nearby Hoboken. Further, only the year 2018 is recorded - cyclical or seasonal patterns may therefore not be fully appreciated.


2. Data cleaning

Preparing data for visualisation (1.7)


Libraries Used:

library(tidyverse)
## -- Attaching packages --------------------------------------- tidyverse 1.3.1 --
## v ggplot2 3.3.5     v purrr   0.3.4
## v tibble  3.1.6     v dplyr   1.0.8
## v tidyr   1.2.0     v stringr 1.4.0
## v readr   2.1.2     v forcats 0.5.1
## -- Conflicts ------------------------------------------ tidyverse_conflicts() --
## x dplyr::filter() masks stats::filter()
## x dplyr::lag()    masks stats::lag()
library(tsibbledata)
## Warning: package 'tsibbledata' was built under R version 4.1.3
library(tsibble)
## Warning: package 'tsibble' was built under R version 4.1.3
## 
## Attaching package: 'tsibble'
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, union
library(lubridate)
## Warning: package 'lubridate' was built under R version 4.1.3
## 
## Attaching package: 'lubridate'
## The following object is masked from 'package:tsibble':
## 
##     interval
## The following objects are masked from 'package:base':
## 
##     date, intersect, setdiff, union
library(ggplot2)
library(ggthemes)
## Warning: package 'ggthemes' was built under R version 4.1.3
library(sf) 
## Warning: package 'sf' was built under R version 4.1.3
## Linking to GEOS 3.9.1, GDAL 3.2.1, PROJ 7.2.1; sf_use_s2() is TRUE
library(rgeos)
## Warning: package 'rgeos' was built under R version 4.1.3
## Loading required package: sp
## Warning: package 'sp' was built under R version 4.1.3
## rgeos version: 0.5-9, (SVN revision 684)
##  GEOS runtime version: 3.9.1-CAPI-1.14.2 
##  Please note that rgeos will be retired by the end of 2023,
## plan transition to sf functions using GEOS at your earliest convenience.
##  GEOS using OverlayNG
##  Linking to sp version: 1.4-6 
##  Polygon checking: TRUE
library(rnaturalearth)
## Warning: package 'rnaturalearth' was built under R version 4.1.3
library(rnaturalearthdata)
## Warning: package 'rnaturalearthdata' was built under R version 4.1.3
library(leaflet)
## Warning: package 'leaflet' was built under R version 4.1.3


Load and view the data set:

nyc_bikes_df <- tsibbledata::nyc_bikes
nyc_bikes_df
## # A tsibble: 4,268 x 12 [!] <America/New_York>
## # Key:       bike_id [10]
##    bike_id start_time          stop_time           start_station start_lat
##    <fct>   <dttm>              <dttm>              <fct>             <dbl>
##  1 26301   2018-02-26 19:11:03 2018-02-26 19:15:40 3186               40.7
##  2 26301   2018-02-27 07:52:49 2018-02-27 07:58:13 3203               40.7
##  3 26301   2018-02-27 12:03:27 2018-02-27 12:04:54 3202               40.7
##  4 26301   2018-02-27 13:53:51 2018-02-27 14:21:04 3638               40.7
##  5 26301   2018-02-27 14:30:42 2018-02-27 14:33:11 3638               40.7
##  6 26301   2018-02-27 16:31:04 2018-02-27 16:33:27 3187               40.7
##  7 26301   2018-02-27 17:37:12 2018-02-27 17:40:49 3638               40.7
##  8 26301   2018-02-27 17:49:10 2018-02-27 17:55:03 3639               40.7
##  9 26301   2018-02-27 18:21:57 2018-02-27 18:25:18 3202               40.7
## 10 26301   2018-02-27 22:08:55 2018-02-27 22:28:42 3638               40.7
## # ... with 4,258 more rows, and 7 more variables: start_long <dbl>,
## #   end_station <fct>, end_lat <dbl>, end_long <dbl>, type <fct>,
## #   birth_year <dbl>, gender <fct>
# glimpse(nyc_bikes_df)
# view(nyc_bikes_df)
# There are 4,2688 rows initially but 4,266 after removing two users who are
# recorded as being born in 1880s


Adding columns

  1. Adding columns: start day of the week, month, time
  2. Creating a column that shows the season just in case
  3. Creating a column for age of rider and removing two outliers (born 1880s)
  4. Showing hire duration (2 were hired for 3 days and the rest less than a day. Most were hired for less than 50 mins)
#creating 4 columns: date, and start day, month and hour
nyc_bikes_df <- nyc_bikes_df %>% 
  mutate(date = as_date(start_time), .after = 2,
         start_day = wday(start_time, label = TRUE, abbr = FALSE),
         start_month = month(start_time, label = TRUE, abbr = FALSE),
         start_hour = hour(start_time), minute(start_time))

nyc_bikes_df <- nyc_bikes_df %>% 
  unite(start_hire_time, 
        start_hour, `minute(start_time)`, 
        sep = ":", 
        remove = FALSE)

#adding a season column just in case
nyc_bikes_df <- nyc_bikes_df %>% 
  mutate(season = case_when(
    start_month %in% c("December", "January", "February") ~ "Winter",
    start_month %in% c("March", "April", "May") ~ "Spring",
    start_month %in% c("June", "July", "August") ~ "Summer",
    start_month %in% c("September", "October", "November") ~ "Autumn",
    ), .after = 5)

# showing duration of hire (2 were hired for 3 days and the rest less than a day
# most were hired for less than 50 mins)
nyc_bikes_df <- nyc_bikes_df %>% 
  mutate(hire_duration = stop_time - start_time, # this shows 1.50(= 1min 30sec)
         length_of_hire_rounded = as.period(round(stop_time - start_time)), 
         hours_hired = as.numeric(length_of_hire_rounded, "hours")
  )

#Finding people's ages in 2018 - 2018 was the year the data was collected
nyc_bikes_df <- nyc_bikes_df %>% 
  mutate(age_years = as.numeric(2018 - birth_year))  

# There are two rows where the customer is born in the 1880s. That can not be
# right so I am removing those two rows (neither has a gender either)
nyc_bikes_df <- nyc_bikes_df %>% 
  filter(birth_year > 1900)
nyc_bikes_df
## # A tibble: 4,266 x 23
##    bike_id start_time          date       start_day start_month season
##    <fct>   <dttm>              <date>     <ord>     <ord>       <chr> 
##  1 26301   2018-02-26 19:11:03 2018-02-26 Monday    February    Winter
##  2 26301   2018-02-27 07:52:49 2018-02-27 Tuesday   February    Winter
##  3 26301   2018-02-27 12:03:27 2018-02-27 Tuesday   February    Winter
##  4 26301   2018-02-27 13:53:51 2018-02-27 Tuesday   February    Winter
##  5 26301   2018-02-27 14:30:42 2018-02-27 Tuesday   February    Winter
##  6 26301   2018-02-27 16:31:04 2018-02-27 Tuesday   February    Winter
##  7 26301   2018-02-27 17:37:12 2018-02-27 Tuesday   February    Winter
##  8 26301   2018-02-27 17:49:10 2018-02-27 Tuesday   February    Winter
##  9 26301   2018-02-27 18:21:57 2018-02-27 Tuesday   February    Winter
## 10 26301   2018-02-27 22:08:55 2018-02-27 Tuesday   February    Winter
## # ... with 4,256 more rows, and 17 more variables: start_hire_time <chr>,
## #   start_hour <int>, `minute(start_time)` <int>, stop_time <dttm>,
## #   start_station <fct>, start_lat <dbl>, start_long <dbl>, end_station <fct>,
## #   end_lat <dbl>, end_long <dbl>, type <fct>, birth_year <dbl>, gender <fct>,
## #   hire_duration <drtn>, length_of_hire_rounded <Period>, hours_hired <dbl>,
## #   age_years <dbl>


Looking at subscribers, customers and unknown genders:

  • 3951 are subscribers
  • 315 are customers
  • 267 have gender = “unknown” and I noticed that most are 49 years old (n = 243)
  • 49 year olds:
    • 243/269 have gender unknown (26 have a gender)
    • 89/269 are subscribers, the rest (180) are customers (out of 315 in total)
# 3951 are subscribers 
nyc_bikes_df %>% 
  filter(type == "Subscriber")
## # A tibble: 3,951 x 23
##    bike_id start_time          date       start_day start_month season
##    <fct>   <dttm>              <date>     <ord>     <ord>       <chr> 
##  1 26301   2018-02-26 19:11:03 2018-02-26 Monday    February    Winter
##  2 26301   2018-02-27 07:52:49 2018-02-27 Tuesday   February    Winter
##  3 26301   2018-02-27 12:03:27 2018-02-27 Tuesday   February    Winter
##  4 26301   2018-02-27 13:53:51 2018-02-27 Tuesday   February    Winter
##  5 26301   2018-02-27 14:30:42 2018-02-27 Tuesday   February    Winter
##  6 26301   2018-02-27 16:31:04 2018-02-27 Tuesday   February    Winter
##  7 26301   2018-02-27 17:37:12 2018-02-27 Tuesday   February    Winter
##  8 26301   2018-02-27 17:49:10 2018-02-27 Tuesday   February    Winter
##  9 26301   2018-02-27 18:21:57 2018-02-27 Tuesday   February    Winter
## 10 26301   2018-02-27 22:08:55 2018-02-27 Tuesday   February    Winter
## # ... with 3,941 more rows, and 17 more variables: start_hire_time <chr>,
## #   start_hour <int>, `minute(start_time)` <int>, stop_time <dttm>,
## #   start_station <fct>, start_lat <dbl>, start_long <dbl>, end_station <fct>,
## #   end_lat <dbl>, end_long <dbl>, type <fct>, birth_year <dbl>, gender <fct>,
## #   hire_duration <drtn>, length_of_hire_rounded <Period>, hours_hired <dbl>,
## #   age_years <dbl>
# 315 are customers
nyc_bikes_df %>% 
  filter(type == "Customer")
## # A tibble: 315 x 23
##    bike_id start_time          date       start_day start_month season
##    <fct>   <dttm>              <date>     <ord>     <ord>       <chr> 
##  1 26301   2018-03-11 08:57:02 2018-03-11 Sunday    March       Spring
##  2 26301   2018-03-22 21:18:03 2018-03-22 Thursday  March       Spring
##  3 26301   2018-04-22 13:07:24 2018-04-22 Sunday    April       Spring
##  4 26301   2018-04-22 16:28:54 2018-04-22 Sunday    April       Spring
##  5 26301   2018-05-12 18:02:32 2018-05-12 Saturday  May         Spring
##  6 26301   2018-05-17 22:32:56 2018-05-17 Thursday  May         Spring
##  7 26301   2018-05-20 13:41:58 2018-05-20 Sunday    May         Spring
##  8 26301   2018-05-28 14:14:29 2018-05-28 Monday    May         Spring
##  9 26301   2018-06-05 19:26:19 2018-06-05 Tuesday   June        Summer
## 10 26301   2018-06-07 08:27:48 2018-06-07 Thursday  June        Summer
## # ... with 305 more rows, and 17 more variables: start_hire_time <chr>,
## #   start_hour <int>, `minute(start_time)` <int>, stop_time <dttm>,
## #   start_station <fct>, start_lat <dbl>, start_long <dbl>, end_station <fct>,
## #   end_lat <dbl>, end_long <dbl>, type <fct>, birth_year <dbl>, gender <fct>,
## #   hire_duration <drtn>, length_of_hire_rounded <Period>, hours_hired <dbl>,
## #   age_years <dbl>
# 267 have gender = "unknown" and I noticed that most are 49 years old (n = 243)
nyc_bikes_df %>% 
  filter(gender == "Unknown" & age_years == 49)
## # A tibble: 243 x 23
##    bike_id start_time          date       start_day start_month season
##    <fct>   <dttm>              <date>     <ord>     <ord>       <chr> 
##  1 26301   2018-03-22 21:18:03 2018-03-22 Thursday  March       Spring
##  2 26301   2018-04-22 16:28:54 2018-04-22 Sunday    April       Spring
##  3 26301   2018-05-25 14:11:15 2018-05-25 Friday    May         Spring
##  4 26301   2018-05-28 14:14:29 2018-05-28 Monday    May         Spring
##  5 26301   2018-06-05 19:26:19 2018-06-05 Tuesday   June        Summer
##  6 26301   2018-06-07 12:43:20 2018-06-07 Thursday  June        Summer
##  7 26301   2018-06-08 10:04:27 2018-06-08 Friday    June        Summer
##  8 26301   2018-06-09 17:49:44 2018-06-09 Saturday  June        Summer
##  9 26301   2018-06-16 18:30:02 2018-06-16 Saturday  June        Summer
## 10 26301   2018-06-17 16:22:57 2018-06-17 Sunday    June        Summer
## # ... with 233 more rows, and 17 more variables: start_hire_time <chr>,
## #   start_hour <int>, `minute(start_time)` <int>, stop_time <dttm>,
## #   start_station <fct>, start_lat <dbl>, start_long <dbl>, end_station <fct>,
## #   end_lat <dbl>, end_long <dbl>, type <fct>, birth_year <dbl>, gender <fct>,
## #   hire_duration <drtn>, length_of_hire_rounded <Period>, hours_hired <dbl>,
## #   age_years <dbl>
# 49 year olds:
# 243/269 have gender unknown (26 have a gender).
# 89/269 are subscribers, the rest (180) are customers (out of 315 in total)
age49 <- nyc_bikes_df %>% 
  filter(age_years == 49)


3. Data visualisation

Process and design (2.1, 2.7, 2.8, 2.9)

All visualisations were made in RStudio using the ggplot package. Having picked up on possible areas of interest during the wrangling process, I investigated patterns of bike hire over time simply by looking at distribution plots of the days of the year, days of the week, and differences in months. I then considered demographics such as age and gender as well as looking into the proportion of “customers” against “subsribers”. Finally, geographical consideration was given to the spread of the start points of bike hire and returns. Having selected key visualisations, I considered their aesthetics and degree of accuracy with the hope of conveying appropriate messages and providing insight in the business context.


Visualisations (2.2, 2.3, 2.4, 2.5, 2.6)

Patterns Across Time


Looking at the total hires per day across the year:

count_per_day <- nyc_bikes_df %>% 
  group_by(date) %>% 
  summarise(rentals_per_day = n()) %>% 
  arrange(desc(rentals_per_day))

count_per_day %>% 
  ggplot()+
  aes(x = date, 
      y = rentals_per_day)+
  geom_line() + 
  labs(x = "\nDays in 2018",
       y = "Total number of rentals per day",
       title = "Total number of bike rentals per day",
       subtitle = "For every day in the year 2018")+
  theme_bw()


From the line graph above, we can see the variation in total number of bike rentals per day across the year 2018.

An increase in total rentals per day can be seen as the graph moves from winter into summer (the highest spike is on 31st July, a Tuesday) and then total hires decrease again as we head into autumn and winter again.


Though a trend can be seen, it is a bit messy. To look at it another way we can look at the total bikes rented per month:

count_per_month <- nyc_bikes_df %>% 
  group_by(season, start_month) %>% 
  summarise(rentals_per_month = n()) %>% 
  arrange(desc(rentals_per_month))
## `summarise()` has grouped output by 'season'. You can override using the
## `.groups` argument.
count_per_month %>% 
  ggplot()+
  aes(x = start_month, 
      y = rentals_per_month) +
  geom_col(fill = "#007cb9")+
  theme_bw()+
  theme(axis.text.x = element_text(angle = 45, 
                                   hjust = 0.9))+
  geom_text(aes(label = rentals_per_month),
    position = position_dodge(0.9),
    vjust = 0)+ 
  labs(x = "\nMonth (in 2018)", 
       y = "Total rentals per month", 
       title = "Total Bike Rentals per Month in 2018")


Following the same pattern as the previous line graph, this graph shows an increase in total bike rentals per month from winter into summer with the most rentals in August. Following on from this, there is a sharp decrease as autumn and winter arrive.

The business may wish to consider furthering these two graphs by looking into seasonal and yearly cycles so as to more accurately predict customer behaviour throughout the year and by season.


With an interest in considering patterns of bike hire over days of the week, the below graph shows the days of the week, also the day individual bikes were hired, against the total number of bikes hired on each day throughout the year:

weekdays <- c("Monday", "Tuesday", "Wednesday", "Thursday", 
              "Friday", "Saturday", "Sunday")

#Tues: 720!, Fri 658, Wed, Thurs, Mon 628, Sat 523, Sun(453!))
bikes_per_day <- nyc_bikes_df %>% 
  group_by(start_day) %>% 
  summarise(num_of_bikes_hired = n()) %>% 
  arrange(desc(num_of_bikes_hired))

bikes_per_day %>%  
  ggplot(aes(x = start_day, 
             y = num_of_bikes_hired))+
  scale_x_discrete(limits = weekdays) +
    geom_col(fill = "#007cb9")+
  theme(axis.text.x = element_text(angle = 45, 
                                   hjust = 0.9))+
  geom_text(aes(label = num_of_bikes_hired),
    position = position_dodge(0.9),
    vjust = 0) +
  labs(
    x = "\nDay of hire", 
    y = "Total number of bikes hired throughout the year", 
    title = "Total number of bikes hired by weekday in 2018"
  )


Here we can see that, overall, weekdays are more popular for bike hire than weekends. 720 bikes were hired on Tuesdays compared with 453 on Sundays. Was this the case every month?

It would be interesting to see if this is a pattern across the year as it may point to customer usage, i.e. are more customers commuters or hiring for leisure:

monthly_bikes_per_day <-nyc_bikes_df %>% 
  group_by(start_month, start_day) %>% 
  summarise(num_of_bikes_hired = n())
## `summarise()` has grouped output by 'start_month'. You can override using the
## `.groups` argument.
monthly_bikes_per_day %>% 
  ggplot()+
  aes(x = start_day, 
      y = num_of_bikes_hired, 
      fill = start_day)+
  scale_x_discrete(limits = weekdays) +
  geom_col(fill = "#007cb9") + 
  facet_wrap(~start_month)+
  theme_bw()+
  theme(axis.text.x = element_text(angle = 45, 
                                   hjust = 0.9))+
  labs(
    x = "\nDay of hire", 
    y = "Total number of bikes hired", 
    title = "Total number of bikes hired by day in each month in 2018")


Though not terribly easy to decipher, it is clear that almost all months show that weekends see similar or fewer bike hires than weekdays throughout the year.

Further research could be undertaken to try and determine the type of user hiring Citi Bikes, possibly by asking customers their main purpose for signing up when they do so.


With the information we have on start times, we can start to gather ideas as to what kind of users sign up to Citi Bike. Are people mostly travelling at commuting times? Let’s look at the full year:

time_of_ride <- nyc_bikes_df %>% 
  group_by(start_hour) %>% 
  summarise(count = n()) %>% 
  arrange(desc(count))

time_of_ride %>% 
  ggplot()+
  aes(x = start_hour, 
      y = count)+
  geom_col(fill = "#007cb9")+
  theme_bw()+
  geom_text(aes(label = count),
    position = position_dodge(0.9),
    vjust = 0)+
  scale_x_continuous(breaks = c(0, 2, 4, 6, 8, 10, 12, 
                                14, 16, 18, 20, 22, 24)) + 
  labs(x = "\nHire Start Time (24hr clock)", 
       y = "Total number of hires", 
       title = "Total number of hires by hour across the year 2018")


And again, seeing if this is a pattern across the months of the year:

#month
time_of_ride_by_month <- nyc_bikes_df %>% 
  group_by(start_month, start_hour) %>% 
  summarise(count = n()) %>% 
  arrange(desc(count))
## `summarise()` has grouped output by 'start_month'. You can override using the
## `.groups` argument.
time_of_ride_by_month %>% 
  ggplot()+
  aes(x = start_hour, 
      y = count)+
  geom_col(fill = "#007cb9")+
  facet_wrap(~start_month)+ 
  theme_bw() +  
  scale_x_continuous(breaks = c(0, 4, 8, 12, 
                                16, 20, 24)) + 
  labs(x = "\nHire Start Time (24hr clock)", 
       y = "Total number of hires", 
       title = "Total number of hires by hour across months of the year 2018")


Without exception, we can see that morning and afternoon/evening peak commuting hours are the most popular for bike hire in each month of the year. Again, if a reason for hire was given by customers, this could be further looked into and marketing strategies could be influenced by such information or the company could look into whether they have the appropriate number of bikes at each station.


Patterns Across Demographics

Gender

  • There are 3,069 males, 930 females, 267 unknown (243 of them are “49” years old)
  • There are 3,999 rows when taking out gender unknown (as described in Data Quality and Data Bias (1.6))
nyc_bikes_df %>% 
  group_by(gender) %>% 
  summarise(count = n())
## # A tibble: 3 x 2
##   gender  count
##   <fct>   <int>
## 1 Unknown   267
## 2 Male     3069
## 3 Female    930
minus_gender_unknown <- nyc_bikes_df %>% 
  filter(gender %in% c("Male", "Female"))


Looking at total rentals per month based on gender: 

rentals_by_month_and_gender <- minus_gender_unknown %>% 
  group_by(start_month, gender) %>% 
  summarise(rentals_per_month = n()) 
## `summarise()` has grouped output by 'start_month'. You can override using the
## `.groups` argument.
rentals_by_month_and_gender %>% 
  ggplot()+
  aes(x = start_month, 
      y = rentals_per_month, 
      group = gender, 
      colour = gender)+
  geom_point()+
  geom_line() +
  scale_color_manual(values = c("Male" = "#007cb9", "Female" = "#ED2E14"))+
  theme(axis.text.x = element_text(angle = 45, 
                                   hjust = 0.9))+
  geom_label(aes(label = rentals_per_month),
    position = position_dodge(1),
    vjust = 0) +
  labs(
    x = "\nMonth in 2018", 
    y = "Total rentals", 
    title = "Total rentals per month in 2018 based on gender", 
    colour = "Gender"
  )


Firstly, there are far more males than females using Citi Bikes, again this information could be considered by the business in order to concentrate promotions on the most appropriate demographic and at the more appropriate time of year.

Males and females follow a similar pattern as rentals are seen to increase as the months move from January towards August, and then decrease again as we enter the colder autumn and winter months.


Further research of the data also showed that male and female patterns were similar between months of the year and starting stations. However, as can be seen below, where males tend to use Citi Bikes more on weekdays, females have a more consistent usage across the week (though if there were more female customers, these patterns may have started to reflect each other):

day_by_gender <- minus_gender_unknown %>% 
  group_by(start_day, gender) %>% 
  summarise(daily_users = n())
## `summarise()` has grouped output by 'start_day'. You can override using the
## `.groups` argument.
day_by_gender %>% 
  ggplot(aes(x = start_day, 
             y = daily_users, 
             fill = gender))+
  scale_x_discrete(limits = weekdays) +
  geom_col(position = "dodge")  +
  scale_fill_manual(values = c("Male" = "#007cb9", "Female" = "#ED2E14"))+
  theme(axis.text.x = element_text(angle = 45, 
                                   hjust = 0.9))+
  geom_text(aes(label = daily_users),
    position = position_dodge(0.9),
    vjust = 0)+ 
  labs(
    x = "Day of hire",
    y = "Total number of hires",
    title = "Total number of bike hires by gender across days of the week in
    the year 2018", 
    fill = "Gender"
  )


Age

Customers range from 19 to 71 years old, the below graph shows the total number of bikes hired per age in years in 2018.

minus_gender_unknown %>% 
  group_by(age_years, gender) %>% 
  ggplot()+
  aes(x = age_years, 
      fill = gender) +
  geom_bar()  +
  scale_fill_manual(values = c("Male" = "#007cb9", "Female" = "#ED2E14"))+
  labs(
    x = "\nAge in years",
    y = "Total number of Citi Bike customers",
    title = "Total number of Citi Bike customers of each age (in years)",
    subtitle = "Separated by gender", 
    fill = "Gender"
  )


Citi Bikes could use this information to consider where best to promote their service and to see if they need to add more bikes to any locations to cater to the age demographics as they are.

As a large proportion of customers are in the 25-35 year bracket, specific media channels and venue advertisements could be considered along with targeted offers.


Georgraphical Spread

# how many bikes from each station (52 start stations): 
bikes_per_start_station <- nyc_bikes_df %>% 
  group_by(start_long, start_lat, start_station) %>% 
  summarise(total_hires = n(), 
            average_duration = round(mean(hire_duration))) %>% 
  arrange(desc(total_hires))
## `summarise()` has grouped output by 'start_long', 'start_lat'. You can override
## using the `.groups` argument.
# 55 end stations
bikes_per_end_station<- nyc_bikes_df %>%
  group_by(end_long, end_lat, end_station) %>% 
  summarise(total_returns = n(), 
            average_duration = round(mean(hire_duration))) %>% 
  arrange(desc(total_returns))
## `summarise()` has grouped output by 'end_long', 'end_lat'. You can override
## using the `.groups` argument.


There are 52 different stations recorded as start stations, and 55 as return stations.

The two visualisations below show each station with a radius circle representing how often the station is used. The wider the radius, the more popular the station. Clicking on each station will give information regarding the station code, the total number of hires (or returns, depending on which you are looking at) from that station, and the average journey length in minutes.


Starting Stations

bikes_per_start_station %>% 
  leaflet() %>% 
  addTiles() %>% 
  addCircleMarkers(lng = ~start_long, 
                   lat = ~start_lat,
                   radius = ~total_hires/10,
                   popup = ~paste0("Start Station: ", start_station, 
                                   "<br> Total Hires: ", total_hires, 
                                   "<br> Average Duration: ", average_duration, 
                                   " mins"))


Return Stations

bikes_per_end_station%>% 
  leaflet() %>% 
  addTiles() %>% 
  addCircleMarkers(lng = ~end_long, 
                   lat = ~end_lat,
                   radius = ~total_returns/10,
                   popup = ~paste0("End Station: ", end_station,
                                   "<br>Total Returns ", total_returns, 
                                   "<br> Average Duration: ", average_duration, 
                                   " mins"))


The Most Popular Starting Stations:

These are also the most popular return stations however the return order is: 1, 4, 2, 3, 5.

  1. 3186 - Grove Street Rail Station
    This is the bustling downtown area with trains leaving for and arriving from the Newark Financial District to the West and New York City to the East.

  2. 3203 - Hamilton Park A tranquil and green neighborhood in the Downtown area.

  3. 3195 - Journal Square Busy business district, residential area, and transport hub (9 min cycle from Grove Street Station as can be seen in the picture below).

  4. 3183 - Exchange Place The Jersey City Financial District know as “Wall Street West”, this area is also busy with ferries leaving and entering surrounding piers across the Hudson River.

  5. 3203 - Newport A mixed use community area.


Journey time between Journal Square and Grove Street


Duration

The analysed data could potentially be pointing towards commuters using the service more than casual customers.

When considering total journey time below, we can see that the majority of journeys are complete within 10 minutes.


duration <- nyc_bikes_df %>% 
  group_by(length_of_hire_rounded) %>% 
  summarise(count = n()) %>% 
  arrange(desc(count)) 

a <- nyc_bikes_df %>% 
  mutate(duration_block = case_when(
    length_of_hire_rounded <= "5M 0S" ~ "0 - 5", 
    length_of_hire_rounded <= "10M 0S" & 
      length_of_hire_rounded > "5M 0S" ~ "5 - 10",
    length_of_hire_rounded <= "15M 0S" & 
      length_of_hire_rounded > "10M 0S" ~ "10 - 15",
    length_of_hire_rounded <= "20M 0S" & 
      length_of_hire_rounded > "15M 0S" ~ "15 - 20",
    length_of_hire_rounded <= "25M 0S" & 
      length_of_hire_rounded > "20M 0S" ~ "20 - 25",
    length_of_hire_rounded <= "30M 0S" & 
      length_of_hire_rounded > "25M 0S" ~ "25 - 30",
    length_of_hire_rounded <= "35M 0S" & 
      length_of_hire_rounded > "30M 0S" ~ "30 - 35",
    length_of_hire_rounded <= "40M 0S" & 
      length_of_hire_rounded > "35M 0S" ~ "35 - 40",
    length_of_hire_rounded <= "45M 0S" & 
      length_of_hire_rounded > "40M 0S" ~ "40 - 45",
    length_of_hire_rounded <= "50M 0S" & 
      length_of_hire_rounded > "45M 0S" ~ "45 - 50",
    length_of_hire_rounded <= "55M 0S" & 
      length_of_hire_rounded > "50M 0S" ~ "50 - 55",
    length_of_hire_rounded <= "60M 0S" & 
      length_of_hire_rounded > "55M 0S" ~ "55 - 60",
    length_of_hire_rounded <= "3H 0S" & 
      length_of_hire_rounded > "1H 0S" ~ "1Hrs - 3Hrs",
    length_of_hire_rounded > "3H 0S" ~ "3Hrs plus",
  ))


durations <- a %>% 
  group_by(duration_block) %>% 
  summarise(count = n()) 

time_blocks <- c("0 - 5", "5 - 10", "10 - 15", "15 - 20", "20 - 25", "25 - 30",
                 "30 - 35", "35 - 40", "40 - 45", "45 - 50", "50 - 55", 
                 "55 - 60", "1Hrs - 3Hrs", "3Hrs plus")

durations %>%   
  ggplot()+
  aes(x = duration_block, 
      y = count) +
  scale_x_discrete(limits = time_blocks) +
  geom_col(fill = "#007cb9")+
  theme_bw()+
  theme(axis.text.x = element_text(angle = 45, 
                                   hjust = 0.9))+
  labs(
    x = "Length of Trip (mins)",
    y = "Total number of hires",
    title = "Total number of bike hires per journey time 
    in blocks of 5 minute intervals",
  )+
    geom_text(aes(label = count),
    position = position_dodge(0.9),
    vjust = 0)


Citi Bikes are clearly very popular for short journeys across Jersey City. Though, of course, more information needs to be gathered, it may be considered that the bikes are used for commuting purposes, maybe for shopping trips, or for people visiting the Downtown area for the evening.


Conclusions

  • The busiest periods are in the summer months and at peak commuting times every month of the year.

  • More often than not, weekdays are more popular than weekends - though is this the case with female customers? More female records may need to be gathered.

  • The age range is 19-71 however 25-35 year olds appear to hire bikes more often than others.

  • Most journeys are short, very often less than 10 minutes, and the busiest rental and drop off stations appear to be in the business, financial, and transport hub areas of Jersey City.

  • All of this could guide the business into maintaining or improving upon marketing strategies - promoting to the right people at the right time and in the right place.

  • They could also look at seasonal and cyclical patterns over a longer period to better predict customer trends.

  • They may wish to look into the data further to consider the volume of bikes available at their most and least popular stations.